And that was last year, this year Covid19 sent online traffic soaring sky-high. Ecommerce brands are seeing revenues soar like never before. But with that follows a downward trajectory of customer experience and higher frustrations, longer response time, and a growing impatience with support services overall. With chatbot-usage increasing by 92.5% from 2019 to 2020, and great customer experience becoming critical, chatbots have gone from “good-to-have” to “must-haves”. They are now something customer demand.

Among the many things that the COVID-19 pandemic has changed for the worse, customer service comes out at the top of customers’ frustrations. Ask anyone who has tried to contact a retailer for a product to return or an airline for changing travel plans.

As cities after cities across Europe and the US started shuttering physical spaces and businesses, there were the sudden spikes that brands started noticing on their websites. The spikes stayed and kept growing three folds, weeks after weeks, and months after months.

A report from {{source}} states that retail platforms have undergone an unprecedented global traffic increase between January 2019 and June 2020, surpassing even holiday season traffic peaks. Overall, retail websites generated almost 22 billion visits in June 2020, up from 16.07 billion global visits in January 2020.

This growth in business comes with its fair share of problems. The biggest one being the pressure that some departments, in particular, have to face. One of the hardest-hit departments within the retail industry due to the pandemic’s unexpected “retail boom” are Customer Support Teams.

The Support Struggle

Struggling with backlogs of unanswered queries, a downward spiral of deteriorating services with Customer Satisfaction Scores (CSAT) going completely out of the window. A survey in the US found that some companies saw hold times balloon by as much as 34 percent and escalations (calls sent up the chain of command) skyrocket more than 68 percent.

The growing customers and online orders have inundated the support teams with a mountain of unanswered tickets and frustrated customers. The study last year in the Harvard Business Review found that the average American consumer spends 13 hours a year stuck on hold trying to resolve problems.

Thirteen hours.

But technology is once again proving to provide a quick and effective solution. Chatbots or Digital Assistants are acting as the tool that serves the many in need within an organization. Chatbots have emerged as a quick, easy, and immediate solution for many of the retail Industry’s support problems. From basic bots that can answer FAQs (Frequently asked questions) to complicated voicebots that can handle the entire refund process on IVR, there are many opportunities to use A.I. and make customer experience a happy place again.

Chatbots for Retail within Service can be used for multiple use-cases. But often the first step is to identify the biggest pain point and then scale from there.

FAQ Bot

If your organization has never used a chatbot before this is the right place to start. A quick FAQ Bot that answers all the questions which are already in the knowledge base. These bots could either just send the link to the right FAQ Article to the customer or go even further at answering them conversationally (using Conversational Design) Typically an FAQ Bot when build well can answer anywhere up to 40% of the queries that come in.

Delivery Status Bot

Customers are usually impatient when waiting for their orders to arrive. A package tracking chatbot can check the status of their delivery and give them the most updated information. BotSupply’s chatbot platform enables connecting to any third-party logistics service with 1-click.

Do it all Fully integrated Bot

Now this one really does it all! Once an organization starts seeing results with bots, it’s usually time to scale up and make the whole business process more efficient. This can be done with a chatbot that not only integrated with logistics providers or answer FAQs but a chatbot that is connected to your CRM, which recognized your customers, address them based on your segmentation. One step ahead and you could connect these bots to your Analytics Dashboard, and that’s where the magic happens. You get insights into your customer’s pain, their moments of joy.

A Gartner study, published recently points out that organizations that harness the power of AI into their customer engagement strategy will see an increase in operational efficiency of 25 percent by 2025. Strategic implementation of chatbots to create unforgettable customer experiences now will be the key differentiator for retailers coming out of the pandemic.

Thirteen Hours: The Average hold time on support calls every American spends every year. was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

In a call with an HR Manager last month from one of the UK’s biggest grocery chain, we asked the question,

“So, how many of these questions that your employees are asking can be answered from inside the FAQs?”

The HR Manager didn’t hesitate a second before responding.

“At least 70%”

Build a business case

The technology underpinning chatbots can create 2 distinct types of solutions, often best used as Stage 1 and Stage 2 in sequential order. The first one, an easy, quick chatbot that answers all the questions answers to which already exist in the organization’s knowledge base, FAQs, policies, information about COVID, etc. This frees up teams to tackle more complex questions and keeps information up to date. The latter is extremely important in a pandemic where information and advice are changing fast.

The second is breed is sophisticated chatbots that mine data that they are fed to provide dynamic responses. These chatbots connected to an organization’s backend HR software systems like Oracle Human Capital Management( HCM) or Workday, PeopleSoft whether on-premises or on the cloud. They provide a truly “intelligent” support: understanding employees’ intentions, fetching data from the backend, and giving the right answers.

Starting small

When an organization has not tried the chatbot technology before, it is imperative to start small and build a business case and get all stakeholders on the same page. This is why an FAQ bot is the best place to start.

It easily automates all questions like:

When can I come to the office again?

Where is my payslip for August?

How many leaves do I have left?

These FAQ bots however simple, act as a mighty 1st line of support digital agents. They respond to employee queries 24/7 and give a sense of immediate resolution.

Grow and scale

The next step is to give a real brain to your bot by connecting it to your systems. Here is where the real fun happens. The bot can personalize conversations, understand and measure employee engagement, morale, and give the HR Team insights into the exact areas to work upon.

Regardless of breed, chatbots have proved their worth during the pandemic. They have enabled the HR department to engage with employees to avoid lengthy wait times for straightforward queries.

Chatbots are here to stay, and we will see smarter, more responsive chatbots appearing, exploiting more sophisticated AI in the near future.

So where can you actually use an HR chatbot?

With entire workforces away from offices and working from home, technology like chatbots helps HR Managers stay connected to their employees. In the post-pandemic world, chatbots can help HR departments in the recruitment and on-boarding processes.

Recruitment Bots

Recruitment bots are almost like a 24/7 assistant to a Hiring Manager. They can help candidates in the Hiring process by guiding them through it.

All those emails and calls about “Wanted to check my application status” can be fully automated and handled by chatbots in a personalized way on-demand.

Onboarding Chatbots

Onboarding bots can engage with candidates and prepare them better for a career at the organization.

They can simplify processes like finding vacancies, sending reminders to candidates, and ensuring a good candidate is never misses a relevant vacancy.

Of course, chatbots will not replace the human touch in HR. On the contrary. They will enable HR teams to focus far more on complex resourcing issues in companies and better assist employees in their career paths.

BotSupply is hosting a series of Webinars focusing on the solutions that chatbots can provide to HR Teams within the Retail Industry.

Join us in the conversation and register for “The Future of HR Series”

The Pandemic’s unexpected toll on HR Teams was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

Let’s face it, these are unprecedented times as we are facing the biggest crisis since WWII.

As the world struggles with COVID-19, businesses and public organizations around the world are also struggling in providing the same quality of support to both their customers and their citizens.

And both businesses and public organizations are recognizing that two major shifts are happening: 1) an acceleration of the remote-first working culture and 2) the need for high-quality support automation.

While we are no experts in 1), even though 65% of our team has always been remote and therefore we face no disruption in the way we work internally and provide our services to our customers and partners, we have something to tell about 2).

In a time of social distancing, as they call it, Conversational AI is an immensely powerful technological tool to help businesses and public organizations to face the current disruption. And they need it more than ever.

Those who invested early-on in Conversational AI are facing minor disruptions, or no disruptions at all, in the way they provide support to their users. By using AI to set up self-service capabilities, indeed, businesses don’t have to worry about putting employees in harm’s way and can still provide high-quality, fast and 24/7 available support.

If you are a business that relies heavily on human-driven customer support operations, we understand these are hard times for you.

Financial institutions like JPMorgan, for example, had to close up to 1,000 branches in the US alone due to COVID-19.

“With social distancing used more and more to limit infections, call centers could potentially be severely understaffed or even closed in the near future,” states Forrester Research.

A survey of 1,000 consumers by Lightico found that 82% are concerned about visiting their local branch.

However, there is a solution to this overwhelming solution and it is called Conversational AI.

High-quality Conversational AI allows you to deploy digital assistants to provide human-like customer support to your customers at scale, not worrying about scaling up your team in a situation like this.

By investing in AI technology that understands the complexities of natural human language across different languages, you can still let your customers ask questions like “Can I get a loan these days?”, “How do I know if I have the virus?”, “What to do if I have the virus?”, depending on the service you provide. Automatically and with no human agents needed.

In an era of social distancing, automation and self-service solutions help businesses serve the needs of their customers while also keeping their employees safe.

At BotSupply, we are committed to spreading the use of Conversational AI with the goal of making it easy for humans to interact with the services they need in a safe, immediate and simple way.

For this reason, we are providing our platform FOR FREE for the next 2 months to new customers investing in Conversational AI to face these unprecedented times.

From the BotSupply team, #staysafe and #stayhome!

Get in touch at francesco@botsupply.ai for any query.

AI in a time of social need was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

INTRODUCTION

Machine-learning technology powers many aspects of modern society: from
web searches to content filtering on social networks to recommendations on e-commerce websites, and it is increasingly present in consumer products such as cameras and smartphones. Machine-learning systems are used to identify objects in images, transcribe speech into text, match news items, posts or products with users’ interests, and select relevant results of search.
Increasingly, these applications make use of a class of techniques called deep
learning.[5]

Deep learning (also known as deep structured learning, hierarchical learning or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high level abstractions in data. In a simple case, you could have two sets of neurons: ones that receive an input signal and ones that send an output signal. When the input layer receives an input it passes on a modified version of the input to the next layer. In a deep network, there are many layers between the input and output (and the layers are not made of neurons but it can help to think of it that way), allowing the algorithm to use multiple processing layers, composed of multiple linear and non-linear transformations.[1][2][3][4][5][6][7][8][9]

Deep Learning has revolutionized the machine learning recently with some of the great works being done in this field. These methods have dramatically
improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. But, the ancient term “Deep Learning” was first introduced to Machine Learning by Dechter (1986)[10], and to Artificial Neural Networks (NNs) by Aizenberg et al (2000)[11]. It was further popularized by the development of Convolutional Networks Architecture by Alex Krizhevsky named ‘AlexNet’ that won the competition of ImageNet in 2012 by defeating all the image processing methods and creating a way for deep learning architectures to be used in Image Processing.[12]

DEEP LEARNING ARCHITECTURE

1. Generative deep architectures, which are intended to characterize the
   high-order correlation properties of the observed or visible data for
   pattern analysis or synthesis purposes, and/or characterize the joint
   statistical distributions of the visible data and their associated classes. In
   the latter case, the use of Bayes rule can turn this type of architecture
   into a discriminative one.
2. Discriminative deep architectures, which are intended to directly provide discriminative power for pattern classification, often by characterizing the posterior distributions of classes conditioned on the visible data; and
3. Hybrid deep architectures, where the goal is discrimination but is
   assisted (often in a significant way) with the outcomes of generative
   architectures via better optimization or/and regularization, or
   discriminative criteria are used to learn the parameters in any of the
   deep generative models in category 1) above. [13]

Despite the complex categorization of the deep learning architectures, the
one’s that are in practice are deep feed-forward networks, Convolution
networks and Recurrent Networks.

DEEP FEED FORWARD NETWORKS

Deep feed-forward networks, also often called feed-forward neural networks, or multilayer perceptrons (MLPs), are the quintessential deep learning models.

The goal of a feed-forward network is to approximate some function f∗. For
example, for a classifier, y = f∗ (x) maps an input x to a category y. A feedforward network defines a mapping y=f(x;θ) and learns the value of the
parameters θ that result in the best function approximation.[1]

In simple terms, the network can be defined as input, hidden and output nodes with data coming in from input nodes, processing is done in hidden nodes and then the output is produced through output nodes. The information flows through the function being evaluated from x, through the intermediate computations used to define f, and finally to the output y. There are no feedback connections in which outputs of the model are fed back into itself and hence the models is called as feed-forward network. The model is shown in Figure [1].

CONVOLUTION NEURAL NETWORKS

In machine learning, a convolutional neural network (CNN, or ConvNet) is a type of feed-forward artificial neural network in which the connectivity pattern between its neurons is inspired by the organization of the animal visual cortex.

Individual cortical neurons respond to stimuli in a restricted region of space
known as the receptive field. The receptive fields of different neurons partially overlap such that they tile the visual field.

The response of an individual neuron to stimuli within its receptive field can be approximated mathematically by a convolution operation.[15] Convolutional networks were inspired by biological processes[16] and are variations of multilayer perceptrons designed to use minimal amounts of preprocessing.[17]

They have wide applications in image and video recognition, recommender
systems[18] and natural language processing.[19]

LeNet was one of the very first convolutional neural networks which helped
propel the field of Deep Learning. This pioneering work by Yann LeCun was
named LeNet5 after many previous successful iterations since the year 1988.
At that time the LeNet architecture was used mainly for character recognition tasks such as reading zip codes, digits, etc.

There are four main components in a ConvNet shown in Figure 2 above:
1. Convolutional Layer
2. Activation Function
3. Pooling Layer
4. Fully Connected Layer

Convolutional layer

Convolutional Layer is based on the term ‘Convolution’, which is a
mathematical operation performed on two variables (f*g) to produce a third
variable. It is similar to cross-correlation. The input to a convolutional layer is a m x m x r image where m is the height and width of the image and r is the number of channels, e.g. an RGB image has r=3. The convolutional layer will have k filters (or kernels) of size n x n x q where n is smaller than the
dimension of the image and q can either be the same as the number of
channels r or smaller and may vary for each kernel. The size of the filters gives rise to the locally connected structure which are each convolved with the image to produce k feature maps of size m−n+1.8 [20]

Activation Function

To implement complex mapping functions, activation functions are needed, that are non-linear in order to bring in the much needed non-linearity property that enables them to approximate any function. Activation functions are also important for squashing the unbounded linearly weighted sum from neurons.
This is important to avoid large values accumulating high up the processing
hierarchy. A lot of activation functions are present that can be used with some of the primarily used ones being sigmoid, tanh and ReLU.

Pooling Layer

Pooling is a sample-based discretization process. The objective is to down-sample an input representation (image, hidden-layer output matrix, etc.), reducing it’s dimensionality and allowing for assumptions to be made about features contained in the sub-regions binned.

This is done to in part to help over-fitting by providing an abstracted form of
the representation. As well, it reduces the computational cost by reducing the number of parameters to learn and provides basic translation invariance to the internal representation.

Some of the most prominently used pooling techniques are Max-Pooling, MinPooling and Average-Pooling.

Fully Connected Layer

The term “Fully Connected” implies that every neuron in the previous layer is connected to every neuron on the next layer. The Fully Connected layer is a traditional Multi Layer Perceptron that uses a softmax activation function or any other similar function in the output layer. [21]

RECURRENT NEURAL NETWORKS

In a traditional neural network we assume that all inputs (and outputs) are
independent of each other. But for many tasks that’s a very bad idea. If you
want to predict the next word in a sentence you better know which words came before it. RNNs are called recurrent because they perform the same task for every element of a sequence, with the output being depended on the previous computations. Another way to think about RNNs is that they have a “memory” which captures information about what has been calculated so far. [22]

A RNN has loops in them that allow information to be carried across neurons while reading in input. In the figure[4], x_t is some input, A is a part of the RNN and h_t is the output. Essentially you can feed in words from the sentence or even characters from a string as x_t and through the RNN it will come up with a h_t.s. Some of the types of RNNs are LSTM, Bidirectional RNNs, GRUs and more.

RNNs can be used in NLP, Machine Translations, Language Modeling, Computer vision, Video Analysis, Image generation, Image captioning, and so on due to the fact that any number of inputs and outputs can be fixed in a RNN making it one to one — many to many model possible. Some of the possible architectures are shown in figure[5] with possible explanation of the models.

APPLICATIONS

There has been a lot of research going on in the field of Deep Learning and a
lot of unique questions are solved using deep learning models. Some of the
best applications of deep learning are:

Colorization of Black and White Images

Deep learning can be used to use the objects and their context within the
photograph to color the image, much like a human operator might approach
the problem. the approach involves the use of very large convolutional neural networks and supervised layers that recreate the image with the addition of color.[24][28]

Machine Translations

Text translation can be performed without any pre-processing of the sequence, allowing the algorithm to learn the dependencies between words and their mapping to a new language. Stacked networks of large LSTM recurrent neural networks are used to perform this translation.[25]

Object Classification and Detection in Photographs

This task requires the classification of objects within a photograph as one of a set of previously known objects.
State-of-the-art results have been achieved on benchmark examples of this
problem using very large convolutional neural networks. A breakthrough in this problem by Alex Krizhevsky et al.[12] results on the ImageNet classification problem called AlexNet.[28]

Automatic Handwriting Generation

This is a task where given a corpus of handwriting examples, generate new
handwriting for a given word or phrase.
The handwriting is provided as a sequence of coordinates used by a pen when the handwriting samples were created. From this corpus the relationship between the pen movement and the letters is learned and new examples can be generated ad hoc.[26][28]

Automatic Game Playing

This is a task where a model learns how to play a computer game based only
on the pixels on the screen.
This very difficult task is the domain of deep reinforcement models and is the breakthrough that DeepMind (now part of Google) is renown for achieving.[27] [28]

Generative Model Chatbots

A sequence to sequence based model was used to create a chatbot which learned to generate it’s own answers when trained on a lot of real live conversational datasets. To know more in detail, visit the link.

Generative Model Chatbots

CONCLUSION

It can be concluded from the article that the deep learning models can be used in a variety of tasks due to their capability of simulating the human brain. A lot of research has been done in the area and a lot of research is going to be done in near future. Although, trust issues are there at the moment, but things will be more clear in the near future.

REFERENCES

1. Ian Goodfellow, Yoshua Bengio, and Aaron Courville (2016). Deep
   Learning. MIT Press. Online
2. Deng, L.; Yu, D. (2014).”Deep Learning: Methods and Applications”(PDF).Foundations and Trends in Signal Processing.7(3–4): 1–199.doi:10.1561/2000000039.
3. Bengio, Yoshua (2009).”Learning Deep Architectures for
   AI”(PDF).Foundations and Trends in Machine Learning.2(1): 1–
   127.doi:10.1561/2200000006.
4. Bengio, Y.; Courville, A.; Vincent, P. (2013). “Representation Learning: A
   Review and New Perspectives”.IEEE Transactions on Pattern Analysis and
   MachineIntelligence.35(8):17981828.arXiv:1206.5538.doi:10.1109/tpami.
   2013.50.
5. Schmidhuber, J. (2015). “Deep Learning in Neural Networks: An
   Overview”.NeuralNetworks.61:85117.arXiv:1404.7828.doi:10.1016/j.neun
   et.2014.09.003.
6. Bengio, Yoshua; LeCun, Yann; Hinton, Geoffrey (2015). “Deep
   Learning”.Nature.521: 436–444.doi:10.1038/nature14539.PMID26017442.
7. Deep Machine Learning — A New Frontier in Artificial Intelligence Research– a survey paper by Itamar Arel, Derek C. Rose, and Thomas P. Karnowski. IEEE Computational Intelligence Magazine, 2013
8. Schmidhuber,Jürgen(2015).”DeepLearning”.Scholarpedia.10(11):32832.d
   oi:10.4249/scholarpedia.32832.
9. Carlos E. Perez.”A Pattern Language for Deep Learning”.
10. R. Dechter (1986), University of California, Computer Science
    Department, Cognitive Systems Laboratory.
11. I. Aizenberg, N.N. Aizenberg, and J. P.L. Vandewalle (2000). Multi-Valued and Universal Binary Neurons: Theory, Learning and Applications. Springer Science & Business Media
12. Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. “Imagenet
    classification with deep convolutional neural networks.”Advances in
    neural information processing systems. 2012.
13. Deng,Li, Three Classes of Deep Learning Architectures and Their
    Applications: A Tutorial Survey, Microsoft Research, Redmond, WA 98052, USA.
14. Feed forward Neural Network
15. “Convolutional Neural Networks (LeNet) — DeepLearning 0.1
    documentation”. DeepLearning 0.1. LISA Lab. Retrieved 31 August 2013.
16. Matusugu, Masakazu; Katsuhiko Mori; Yusuke Mitari; Yuji Kaneda (2003). “Subject independent facial expression recognition with robust face detection using a convolutional neural network”(PDF). Neural Networks. (5): 555–559. doi:10.1016/S0893–6080(03)00115–1.
17. LeNet-5, convolutional neural networks
18. van den Oord, Aaron; Dieleman, Sander; Schrauwen, Benjamin (2013–01- 01). Burges, C. J. C.; Bottou, L.; Welling, M.; Ghahramani, Z.; Weinberger, K. Q., eds. Deep content-based music recommendation (PDF). Curran Associates, Inc. pp.2643–2651
19. Collobert, Ronan; Weston, Jason (2008–01–01). “A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask
    Learning”.Proceedings of the 25th International Conference on Machine
    Learning. ICML ’08. New York, NY, USA: ACM: 160–167.doi:10.1145/1390156.1390177. ISBN 978–1–60558–205–4.
20. CNN
21. Intuitive Explaination ConvNets
22. RECURRENT NEURAL NETWORKS TUTORIAL, PART 1 — INTRODUCTION TO RNNS
23. RNN effectiveness
24. Cheng, Zezhou, Qingxiong Yang, and Bin Sheng. “Deep colorization.”
    Proceedings of the IEEE International Conference on Computer Vision.
    2015.
25. Sutskever, Ilya, Oriol Vinyals, and Quoc V. Le. “Sequence to Sequence
    Learning with Neural Networks.” arXiv preprint arXiv:1409.3215 (2014).
26. Graves, Alex. “Generating sequences with recurrent neural networks.”
    arXiv preprint arXiv:1308.0850 (2013).
27. Mnih, Volodymyr, et al. “Playing atari with deep reinforcement learning.” arXiv preprint arXiv:1312.5602 (2013).
28. Application Deep Learning
29. Generative Model Chatbots
30. Understanding LSTMs

To know about the cool things we do at Botsupply every week, follow the weekly posts by Giovanni Toschi every week.

The Week! — issue 4

Check out my other posts on chatbots here.

Kumar Shridhar - Medium

A Beginners Guide to Deep Learning was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

What if we could create a bot who can speak images and gifs as fluent as English and could recognize images in a live conversation ? This is the story of how we created ImagoBuddy who can recognize household objects like vacuum bags and bulbs.

The background

At BotSupply, we try to challenge the boundaries of applied AI and chatbots working together. In our quest to make smarter chatbots, we at BotSupply have created different kind of bots for many enterprise clients. This time, I with our computer vision team, wanted to explore more on how vision could empower bots, and we came up with ImagoBuddy.

The use case

The use-case could be a very simple one. For instance, we could take a starting point in an everyday situation like changing household products such as lightbulbs, a tap or a vacuum bag. The process is surprisingly troublesome, and if you’ve ever tried taking out your last vacuum bag, you know this.

First, we have find out what kind of product we have. Then, we need to find out where to get it, and how to order or which needed to be replaced. The first issue is the hardest, and this is where both computer vision and a platform agnostic chatbot can help!

What if you could just send a picture to a AI powered bot and it can identify product for you…

Train hard your model and harvest fruits with REST : Ahhhh!

Step 1: Implement the AI

First task was to implement an AI which could recognize objects. Creating such an image recognition system, requires data. We collected this data for household products as videos of each object. We then converted videos to images, and trained on our neural network to identify household objects in real-time. We used transfer learning and fine-tuning techniques with some tweaks and were able to identify objects correctly 9 out of 10 times.

Transfer learning is a technique where one can obtain features from neural network trained on large real-world data like Image-net, and use those features to train on a small data like our case of household products.

You can find a nice blog over how to perform transfer learning from the author of Keras, the famous deep learning library, here:

Building powerful image classification models using very little data

Step 2: Deploy the AI

The next task was an important one, to provide our trained deep learning model as a ready-to-use web API. We deployed our model on Google Cloud Platform. All we needed then was a REST call to access our API. More details on how one could do this can be found on a nice article from Rahul:

Tensorflow + Docker = Production ready AI product

Step 3: Design the chatbot UX

It is a very important task to design the bot well, as its an interface directly being used by the end-users. Designing a chatbot as ImagoBuddy, is something that starts all the way back with defining the use-case and understanding context and users, which our Conversational UX designers Simon and Patrick have described in detail here:

Designing for AI — why & how

ImagoBuddy lives on Facebook Messenger, where we used the Messenger Platform APIs to make our bot accessible anywhere over internet.

Once the bot design is ready, we needed to just make a REST call from our bot to our web-based API whenever needed to get real-time response.

Step 4: Pushing chatbots forward with NLP and Computer-Vision

Finally, we presented our ImagoBuddy at Innovation Roundtable Summit 2017 in Copenhagen, which was well received by all the participants.

Challanges

As nothing comes on easy, we also had technical challenges which we improved in later versions of ImagoBuddy. One of the major challenge was to handle situations when user sends a random image to our chatbot which do not belong to any covered household category. This was tackled by training our neural network classifier for a “none” category.

What’s next?

Our next ambition is to create bots which can do stuffs like sign language reading, detecting objects and understand music. I wrote this small article to give a glimpse of how such a bot was created at Botsupply and how you can create one.

Thanks for going through this article! please write your critics, valuable feedback or queries in comments. If you need another article with more details, please mention so in comments section.

For any other communications, please contact me on saurabh@botsupply.ai

ImagoBuddy, a Chatbot recognizing objects in images was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

We are a team scattered across the globe and the way we work , it does not matter at all. It’s 2017 and geography should be irrelevant, and so it is. We complete 1 year today, 7th of December, and on our first birthday, here I write about my wonderful journey with BotSupply so far.

Beginning of 2017 : I was looking for some antidote for my restless mind and my good friend Rahul Kumar approached me for a freelancing work, which he described as something related to AI. That was enough to get me going. He called it SupplyBot. (His local folder is still named like that 😃). Soon enough I had a call with three people, on the different side of the globe and by the end of the call I was not even able to pronounce their names properly but what kicked was the idea. (The Founding Fathers 😂: Asser Traberg Smidt, Giovanni Toschi, Francesco Stasi).Of all the freelancing works that I have done till now, all of them had a plan and an end result, here it was just an idea that could grow endlessly. I jotted down some keywords on my notepad and stared at it after the call. I was happy because things had just started and I had an opportunity to craft things from the scratch with the team.

Since then there has been no turning back. I have been doing what I love and making some lifetime friends and memories along the way. In the picture below you can see me (with the beard and better smile 😂) playing with my toys (my mac and my guitar) and working on BS AI stuff along with Rahul.

Here is how my first idea of the things looked like on paper

And here is how it gradually took shape after continuous efforts of few people scattered across the globe.

Mid Year 2017 : By July we had a system in place with some amazing bots already in market and a larger and stronger team. The whole dashboard I coded was a quick and dirty hack and had scope of lot of improvements, we decided to let go off it in July for a far better and stronger system which now is live and was created by some of the amazing folks at BS.(Kosvo Firepower ! 🔥. Mendim Mestani, Ekrem and Jura). I moved from dashboard to bots and enjoyed working on a fantastic framework, Mbot, built by another champ(Carlito from Venezuela) in the team. I got to wear all the hats from dashboard to bots to finally AI since September. Moving from one domain to another doesn’t mean leaving the previous behind but adding a new one to the feather. This is how we work at BS, everyone helping everyone, at everything.

This is how we work at BS, everyone helping everyone, at everything.

Its been almost a couple of months for me at the cognitive team and we have released a stable system for our NLP called Pachamama, improved the efficiency and improvised the feedback system. While our dashboard has evolved into a beast with features like analytics and broadcasting in place. And our environment expert Animesh has created remarkable system like Jatana and internal tools like Gogo for monitoring.(No one can beat him in naming things 😂).

Looking back, it appears to be one hell of a year. And I can only be proud of what we have been able to achieve together.

Although I have not met any one except Rahul from the team, it never feels so. We discuss all the stuff from code to comedy on our Slack channels and are always aware of what we all are upto.

We don’t work by the clock, but we work by the work

My work table has changed a lot this year. I have worked from the comfort of my room to all sort of places. From StarBucks and CCD’s to mountains and deserts. From lonely drives to crowded yards , from silent shacks to chaotic junctions and many more.

Mostly I have spent my weekends and evenings sipping coffee and coding for BS all across Delhi, all alone.

Sometimes when Rahul is here, we make sure we speed up things and move around Delhi while do it.

I took this picture below right in the middle of a call with the team, in early June. I bet you, no conference room call can beat this view and place. This is some kilometres above Nainital

and below is where I completed the code for Netto bot and took the team call.

and like this, Drinks, Live Music and talks about AI. It was hell of an experience to optimise the accuracy of our NLP engine and set up feed back system with live music . 😎

I spent a weekend in the hills of Mussoorie to dive deep into AI and know the machines more closely.

Apart from NLP we have a crazy Computer Vision system that powers our bots and makes our bots smart enough to identify objects 😱 Kumar Shridhar and team are building crazy stuff in this domain and are powering our “Visionary” bots.

Our marketing and design is in great hands with Simon Morel, Troels and Grasia taking the lead and team is bigger than ever.

Today: Our NLP beast Pachamama, is smarter, faster and filled with new surprise features. (New announcements soon ! 🍿 🚀 🎉)

Its already a year since we started and we have come a long way . We have made some amazing bots, some crazy AI powered solutions and have a bigger and stronger team. (Many new people on board). This journey has just started and I believe we have a long way to go. I would like to end this with Francesco Stasi’s quote.

The world is full of exciting and meaningful stories, and I feel ours could be one of them too.

PS : To everyone in the team, It’s a pleasure working with all you amazing folks, let’s roll up our sleeves and push for the next level, let’s make history. Cheers to the first year and many more to come 💐 🎉 🔥 💪 👊

How it feels like to work @ BotSupply was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

So now we all know that Capsule Networks (by Geoffrey Hinton) is shaking up the AI space and literature states that it will push the limits of Convolutional Neural Network (CNN) to the next level. There are lot of Medium posts, articles and research papers available that discuss about the theory and how it is better than traditional CNN’s.
So I am not going to cover that part, instead I would try to implement the CpNet on TensorFlow using Google’s amazing internal tool called Colaboratory.

Few links you can follow to understand the theory part of CpNet:

• Geoffrey Hinton talk “What is wrong with convolutional neural nets ?”
• Capsule Networks Are Shaking up AI
• Dynamic Routing Between Capsules

Let’s code the network.

Before starting, you can follow my CoLab Notebook and execute the following code.

CoLab URL : https://goo.gl/43Jvju

Now we will clone the repository and install the dependencies. Then we will take the MNIST dataset from the repository and move it out to the parent directory.

!git clone https://github.com/bourdakos1/capsule-networks.git
!pip install -r capsule-networks/requirements.txt
!touch capsule-networks/__init__.py
!mv capsule-networks capsule
!mv capsule/data/ ./data/
! ls

Let’s import all the modules.

import os
import tensorflow as tf
from tqdm import tqdm

from capsule.config import cfg
from capsule.utils import load_mnist
from capsule.capsNet import CapsNet

Initialise the Capsule Network

capsNet = CapsNet(is_training=cfg.is_training)

This is how Capsule Network (CpNet) looks like on Tensorboard graph.

Training the Capsule Network

tf.logging.info('Graph loaded')
sv = tf.train.Supervisor(graph=capsNet.graph,
                             logdir=cfg.logdir,
                             save_model_secs=0)

path = cfg.results + '/accuracy.csv'
if not os.path.exists(cfg.results):
    os.mkdir(cfg.results)
elif os.path.exists(path):
    os.remove(path)

fd_results = open(path, 'w')
fd_results.write('step,test_acc\n')

And creating TF session and running the epochs.
By default the model will be trained for 50 epochs at a batch size of 128. You are always welcome to change the config and try new combinations of hyper parameters.

with sv.managed_session() as sess:
    num_batch = int(60000 / cfg.batch_size)
    num_test_batch = 10000 // cfg.batch_size
    teX, teY = load_mnist(cfg.dataset, False)
    for epoch in range(cfg.epoch):
        if sv.should_stop():
            break
        for step in tqdm(range(num_batch), total=num_batch, ncols=70, leave=False, unit='b'):
            global_step = sess.run(capsNet.global_step)
            sess.run(capsNet.train_op)

if step % cfg.train_sum_freq == 0:
                _, summary_str = sess.run([capsNet.train_op, capsNet.train_summary])
                sv.summary_writer.add_summary(summary_str, global_step)

if (global_step + 1) % cfg.test_sum_freq == 0:
                test_acc = 0
                for i in range(num_test_batch):
                    start = i * cfg.batch_size
                    end = start + cfg.batch_size
                    test_acc += sess.run(capsNet.batch_accuracy, {capsNet.X: teX[start:end], capsNet.labels: teY[start:end]})
                test_acc = test_acc / (cfg.batch_size * num_test_batch)
                fd_results.write(str(global_step + 1) + ',' + str(test_acc) + '\n')
                fd_results.flush()

if epoch % cfg.save_freq == 0:
            sv.saver.save(sess, cfg.logdir + '/model_epoch_%04d_step_%02d' % (epoch, global_step))
fd_results.close()
tf.logging.info('Training done')

For running 50 epochs of Capsule network it took me around 6 hours on NVIDIA TitanXp card.
But after successful training I was able to achieve 0.0038874 total loss which is incredible. 😃 💯

Download my trained Capsule model

CpNet Model URL : https://goo.gl/DN7SS3

YOU NEVER KNOW UNTIL YOU TRY IT

If you liked this article, my notebook and models, please applause👏 and share with others!

For any queries , reach out to me via LinkedIn , Twitter or email me on rahul@botsupply.ai.

Running CapsuleNet on TensorFlow was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

BotSupply is exited to join the NVIDIA inception program aimed at nurturing startups working with innovative technologies in the field of AI

We are excited to officially join the NVIDIA family!

A couple of weeks ago, BotSupply was indeed selected for the Inception Program.

“Inception nurtures dedicated and exceptional startups who are revolutionizing industries with advances in AI and data science. This virtual accelerator program helps startups during critical stages of product development, prototyping, and deployment.”

We look forward to work side by side with NVIDIA, and together grow and improve our Machine Learning and Deep Learning solutions in the NLP and Computer Vision fields.

Exciting times are ahead of us, stay tuned!

BotSupply joins NVIDIA Inception Program was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

And so we met again! 👋

Chatbots Copenhagen is the community driven meetup organised by BotSupply where chatbot enthusiasts, bot experts and other peeps curious about Chatbots meet up every month.

Chatbots Copenhagen is where all the cool kids hang!😎

The events are characterised by being charged with inspiring knowledge sharing and enthusiastic networking! Like minded individuals come together to share their experience and projects with other members of the community in our different sessions and we close down every time with our traditional “Open Mic” session.

In our Open Mic sessions 🎤🎤🎤:

Members are encouraged to share whatever questions they might have about developing chatbots with the rest of the community.

It’s also possible for companies to demo their new bot related projects and get some feedback on it.

Chatbots Copenhagen obtains incredible value from the fact that its made up by a very diverse crowd, from students to experts in different fields, we’ve touched down in several different relevant topics from design and AI to “breakfasts” bots.

Here’s a run through of our 5th and 6th edition:

• Thomas Martinsen from Bluefragments introduced us to the charming Emma, a bot that orders breakfast for the guys at Bluefragments every week. 🍪 ☕️
• Viktor Georgo introduced us to the efficient and hardworking Atori, a personal assistant that knows the project better than the people working on it. 🙋🏻
• Simon Morel from BotSupply talked to us about “How we can create a design process blueprint for developing AI products”. 🤔
• Eleonora Kurilchik from Wiredelta shared with the crowd their journey and enlightening experience around “10 months of learning how to Chatbot” 🤓
• Oracle showcased and presented to the attendees the launch of their brand new chatbot platform. Oracle DK is getting very active in the local startup community, and we hope to see them engaged in many more activities like this. Thanks again Lars and whole Oracle team for hosting the event! 🙌
• Rahul Kumar from BotSupply presented Pachamama a complete ecosystem around Natural Language Understanding (NLU), that allows our solutions to talk to clients in their native natural languages such as Danish, Italian, English, etc.

As you can see a lot has happened in the last two months. 😃 ⭐️ We’re pretty sure Chatbots Copenhagen is the coolest meetup around!

It’s been quite the exciting ride and tons of talented people have been able to connect with each other through Chatbots Copenhagen in a friendly and affable environment. This allows for the community to continue to grow, with new members joining us every time. We are all about community and sharing knowledge. Our goal is to build a space that helps others in building great solutions and that nurtures exciting projects for the benefit of all. This openess is what makes this community simply awesome!

Plus, there’s always food! On our last event we meet at the headquarters of Oracle Copenhagen where we were treated to delicious, Italian approved, fresh from the oven pizza! 🍕🍕🍕🍕

As you can see, we are pretty proud of what Chatbots Copenhagen has become. And this is only the beginning, we are determined to continue to facilitate spaces where people can learn more about AI and it’s applications. So stay tuned for more!

On November you can find us:

• Teaching people how to build their own chatbot from scratch together with the peeps of CBS Code. You can sign up for the event here. 🤖 (Nov. 16)
• Hosting a special edition of Chatbots Copenhagen, because it has come up several times on our monthly meet ups, we’ve decided to dedicate a whole session to exploring “The impact of AI in Society”. You can sign up here. 🚀 (Nov. 20)

We hope to see you around and thanks for helping us build such a kick-ass community, it’s you that make Chatbots Copenhagen awesome! ⭐️ 🚀⭐️

We always live stream our events, in case you can’t join. Last time 1.5K people joined us through FB live, if you wish to check them out, the recordings are still available on BotSupply’s FB page.

Ps.

If you can't get enough of chatbots and once a month is simply not enough for you, join the dialogue on our FB group: Chatbots Copenhagen to stay updated in all the latest details around the community.

Chatbots Copenhagen: The flourishing Community 🤖 was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.

Initial ponderations on whether your company is ready to make the leap into AI implementation 🚀

In the previous AI Playbook episodes, we’ve taken comprehensive look at AI and it’s implementation. We started by taking a look at the main trends in artificial intelligence, followed by a look at a day in the life a futuristic view of how our daily lives will look like when more and more aspects of our daily routine will be powered by AI. We discussed all the industries that are being reshaped by AI and finally we introduced the Human + bot equation, a profound look at how AI is helping humans to become better at what they do.

We’ve been making the case that business environments will become much more “intelligent”, due to the possibilities unlocked by the human +bot equation which will allow for the augmentation of workers’ capacities by forming a clean bond between machines and humans.

The work place will become more competitive, people will increase their production and will have more time to innovate, thus improving the users experiences’ across sectors.

Sooner or later, AI will have to be implemented in all companies that do not want to lose their competitive position in the market. In a report developed by Accenture, it is estimated that by 2035 those companies that incorporate AI into their business model will see their profits increased by 38%.

Artificial Intelligence is rapidly becoming a fundamental factor for production, it has the potential to introduce new sources of growth because of it’s ability to be highly customisable which will eventually transform work and boost the economy.

But how do you know if your company is ready for AI?

Assessing readiness: aspects to pay attention to when looking into Implementing AI in your company

Aspect 1:

To implement AI in your company you have to start by considering what are the actual applicabilitions of AI and then try to match this application with a business pain or an area your company is trying to improve.

Let’s take a look at the most common applications of AI. (Note: AI applicabilities are not limited to the options presented below, these are just a few of the most common ones)

1. Virtual Assistants:
   Businesses use chatbots for a variety of cases, such as customer service. Simply put, an artificial intelligence service can be used to answer simple questions, help users book services, get more information about a specific topic, buy a product, etc. Having a chatbot help expedite this types of tasks, allows for human agents to focus on more relevant problems. At the same time a chatbot allows the company to have a 24/7 service to attend to their customers needs.
2. Idea Generation:
   Data is the commodity that powers the digital economy these days. However, it is necessary to have the necessary resources to transform them into something of value. Ideally companies will have cognitive solutions in place that learn automatically from all the data they collect. What makes artificial intelligence systems so powerful is precisely the fact that they can learn. That allows them to adapt when market behaviour changes, as well as continuously improve performance as more data comes in.
3. Automation of manual processes:
   Artificial intelligence is rapidly automating routine and mechanical cognitive processes. Leaving more time for innovation. The use of intelligent algorithms, for example, can now automate the process of collecting data from various reports and perform an analysis to determine the profitability of a particular business path.
4. Analysis of unstructured data:
   It is estimated that 80% of the digital data is not structured. Organizing and tracking these data has the potential of leading to a better understanding of the users and making predictions based on tendencies.

Aspect 2:

Evaluate the competition

If you’re looking to boost your perspective around how and when to implement AI. An innovation assessment of what your competitors are doing with AI its a good place to start. Benchmarking has traditionally been used as a tool for evaluation and diagnosis of innovation systems.

For a business to thrive it is essential that it takes measures to remain at the forefront and to remain competitive within the market. In this sense, benchmarking can allow for a company to identify areas of opportunity.

AI powered innovation has the potential of becoming exponential, so if you’re competitors are starting to analyse their big data with cognitive solutions, tending to their clients 24/7 with the help of chatbots, etc. Take this as a definite sign that its time to implement AI in your company.

Aspect 3:

Run a quantitative analysis of the costs and benefits of AI implementation.

Innovation is a force of change for any sector. BUT before jumping in the digitalisation wagon, take a simple yet vital final step. Run the numbers to help you understand how much your company can invest in renewing or updating some processes. Then compare it with the benefits of it, it’s very likely the benefits will by far exceed the costs, but it never hurts to check. Your accounting department will appreciate it!

So what’s next?

Now that you have taken this initial considerations, the next logical thing to do is to start taking action.

For this, at BotSupply we have designed a “Digital Transformation Readiness tool” that will help you evaluate how ready your business is to join the AI Revolution. This tool has been developed by following state of the art research around digital transformation and the impact of AI, and it will help you gain further inights in this ever so relevant theme.

See you next episode!👋

The AI Playbook — part 5: Considerations to make prior immersing on an AI led digital… was originally published in BotSupply on Medium, where people are continuing the conversation by highlighting and responding to this story.